CN111584027A - Brain control rehabilitation system motor imagery recognition system fusing complex network and graph convolution - Google Patents

Brain control rehabilitation system motor imagery recognition system fusing complex network and graph convolution Download PDF

Info

Publication number
CN111584027A
CN111584027A CN202010364649.7A CN202010364649A CN111584027A CN 111584027 A CN111584027 A CN 111584027A CN 202010364649 A CN202010364649 A CN 202010364649A CN 111584027 A CN111584027 A CN 111584027A
Authority
CN
China
Prior art keywords
motor imagery
entropy
eeg
brain
electroencephalogram
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010364649.7A
Other languages
Chinese (zh)
Other versions
CN111584027B (en
Inventor
高忠科
吕冬梅
党伟东
马超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN202010364649.7A priority Critical patent/CN111584027B/en
Publication of CN111584027A publication Critical patent/CN111584027A/en
Application granted granted Critical
Publication of CN111584027B publication Critical patent/CN111584027B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Abstract

A brain control rehabilitation system motor imagery recognition system fusing a complex network and a atlas carries out motor imagery by watching hand fist making and stretching action videos of a testee, and meanwhile, electroencephalogram signal acquisition equipment acquires motor imagery EEG electroencephalogram signals of the testee; the motor intention identification module constructs a multi-entropy complex network for the obtained motor imagery EEG electroencephalogram, extracts the characteristics of the motor imagery EEG in the aspects of symbol fluctuation, frequency energy distribution and amplitude fluctuation, inputs the characteristics into a convolutional neural network to classify and identify the fist-making motor imagery EEG electroencephalogram and the hand stretching motor imagery EEG electroencephalogram, and transmits the classification result to a brain-controlled rehabilitation system to prompt a testee to perform the actions of making a fist and stretching the hand; the brain control rehabilitation system enables a closed loop path to be formed between the movement intention and the body sensation of the limbs, gradually enhances the muscle strength and the nerve conduction speed of a testee, promotes the recovery of the damaged brain movement area, and gradually recovers the activity.

Description

Brain control rehabilitation system motor imagery recognition system fusing complex network and graph convolution
Technical Field
The invention relates to a motor imagery identification system. In particular to a brain control rehabilitation system motor imagery identification system fusing a complex network and a graph volume.
Background
A brain-computer interface (BCI) system provides a connection path between a human brain and an external device, and the system firstly collects brain activity signals, then detects the intention of a user through a signal processing part, and finally converts the intention into instructions to control the external device. The motor thinking is a classical BCI paradigm, when a person imagines limb movement, the person can cause a certain area of a brain motor perception cortex to be activated, the activation of the motor perception cortex can cause the potential change of the cortex, and the electric signal is called as a motor imagination signal. Different motor imagery tasks will observe oscillatory activity in different areas of the sensory motor cortex of the brain. BCI systems using motor imagery have important values in the fields of neuroscience and rehabilitation, and have been used for recovering nerve transmission function of brains of stroke patients and helping the patients to recover impaired motor functions. The rehabilitation system based on motor imagery can convert the motor intention of a patient into limb movement of the patient, and helps the patient to perform active rehabilitation training better. The research on motor imagery mainly consists in classifying and feature extracting the acquired motor imagery EEG signals. The cospatial model (CSP) is a classical method for extracting features in the study of motor imagery, and other feature extraction and dimension reduction methods such as Independent Component Analysis (ICA) and Principal Component Analysis (PCA) are also often used to improve the classification accuracy of motor imagery. In the classification section, many conventional algorithms, such as Support Vector Machines (SVMs) and Linear Discriminant Analysis (LDA), have been widely used. However, because the electroencephalogram signals have the characteristics of weak signals, large noise and the like, the accurate classification of the motor imagery signals cannot be realized at present, and the accurate extraction of the motor imagery signal features needs to be further explored. The complex network is used as an emerging nonlinear complex system analysis tool, and can be used for characterizing and extracting features of a complex system. Graph convolution can process the graph, extract features of higher levels and classify signals. The complex network and the atlas neural network are combined, so that the classification accuracy of the motor imagery signals can be improved, and the applicability of the brain control rehabilitation system is enhanced.
Disclosure of Invention
The invention aims to solve the technical problem of providing a brain-controlled rehabilitation system motor imagery recognition system which can extract and recognize the features of an EEG (electroencephalogram) signal of a fist making motor imagery and an EEG signal of a hand stretching motor imagery and convert the features into a complex network and graph convolution fused with a brain-controlled rehabilitation system controlled by instructions.
The technical scheme adopted by the invention is as follows: a brain control rehabilitation system motor imagery recognition system fusing a complex network and a atlas carries out motor imagery by watching hand fist making and stretching action videos of a testee, and meanwhile, electroencephalogram signal acquisition equipment acquires motor imagery EEG electroencephalogram signals of the testee; the motor intention identification module preprocesses the obtained motor imagery EEG electroencephalogram signals to construct a multi-entropy complex network, the multi-entropy complex network can fuse multi-channel motor imagery EEG electroencephalogram signals and extract the characteristics of the motor imagery EEG electroencephalogram signals in the aspects of symbol fluctuation, frequency energy distribution and amplitude fluctuation; the motor intention identification module inputs an adjacent matrix of the multi-entropy complex network into the atlas neural network to classify and identify the fist making motor imagery EEG electroencephalogram and the hand stretching motor imagery EEG electroencephalogram, and transmits the classification result to the brain control rehabilitation system to prompt the testee to execute the actions of making a fist and stretching the hand; the brain control rehabilitation system enables a closed-loop path to be formed between the movement intention and the body sensation of the limbs, gradually enhances the muscle strength and the nerve conduction speed of a testee, promotes the recovery of the movement area of the damaged brain, and finally realizes that the patient obtains the damaged movement function again through the autonomous training.
The brain-controlled rehabilitation system comprises: the upper limb rehabilitation training device comprises a main controller connected with the movement intention recognition module, and a myoelectric signal acquisition and multi-channel electrical stimulation output module connected with the main controller, wherein the myoelectric signal acquisition and multi-channel electrical stimulation output module applies electrical stimulation to the upper limb of a tested person according to a control instruction of the main controller, feeds the electrically stimulated myoelectric signal back to the main controller, and the main controller generates a control instruction according to a classification result received from the movement intention recognition module and the electrically stimulated myoelectric signal, controls the myoelectric signal acquisition and multi-channel electrical stimulation output module to stimulate the upper limb of the tested person, and helps the tested person to perform upper limb rehabilitation training.
The motion intention identification module specifically comprises the following steps:
1) the method comprises the steps that a testee respectively watches a hand fist making video and a hand stretching video, meanwhile, motor imagery is carried out on corresponding actions of the videos, electroencephalogram signal acquisition equipment is used for acquiring an EEG (electroencephalogram) electroencephalogram signal of the testee for making a fist making motor imagery and an EEG electroencephalogram signal of the hand stretching motor imagery, the EEG electroencephalogram signals are collectively called as the EEG electroencephalogram signals of the motor imagery, and the EEG electroencephalogram signals of the motor imagery are preprocessed;
2) respectively calculating weighted permutation entropy S for the preprocessed motor imagery EEG signals of each electrode1Wavelet packet energy entropy S2Sum amplitude entropy S3Three entropies, and the three entropies are normalized based on a min-max standardization method;
3) subjecting the weighted permutation entropy S of the step 2)1Wavelet packet energy entropy S2Sum amplitude entropy S3These three entropies are constructed into a feature vector [ S ]1,S2,S3]；
4) Constructing a multi-entropy complex network based on the feature vectors in the step 3);
5) respectively constructing a multi-entropy complex network for a fist-making motor imagery EEG electroencephalogram signal and a hand stretching motor imagery EEG electroencephalogram signal in each motor imagery EEG electroencephalogram signal of each testee, and sending an adjacent matrix of the multi-entropy complex network and a classification label thereof into a graph convolution neural network for feature learning and classification;
6) and transmitting the classification result to a brain control rehabilitation system to stimulate the upper limbs of the testee.
The collecting of the EEG signals of the fist making motor imagery and the hand stretching motor imagery of the testee in the step 1) is to collect the EEG signals of the motor imagery of nine electrodes C3, C4, F3, F4, P3, P4, T7, T8 and Cz through brain electrode caps distributed according to 10-20 international standard leads, and carry out preprocessing, wherein the preprocessing is to carry out artifact removal, 8-30Hz band-pass filtering and averaging on the EEG signals of the motor imagery to obtain the EEG signals capable of being used for realizing motor imagery state identification:
Figure BDA0002476162310000021
wherein Xc,iAnd the number of sampling points in channels corresponding to the electrodes with equal length is L.
Separately calculating a weighted permutation entropy S as described in step 2)1Wavelet packet energy entropy S2Sum amplitude entropy S3The three entropies are specifically as follows:
(1) weighted permutation entropy S1The calculation method of (2) is as follows: motor imagery EEG electroencephalogram
Figure BDA0002476162310000022
The time delay embedding expression of (1) is:
Figure BDA0002476162310000023
wherein, Xc,iRepresents the ith sampling point in the motor imagery EEG electroencephalogram signal collected by the c electrode, L represents the number of sampling points in the channel corresponding to the c electrode,
Figure BDA0002476162310000024
representing the u-th phase space vector generated by the channel corresponding to the c electrode, d is an embedding dimension, d is 4, tau is a delay time, tau is 1, and L- (d-1) tau represents an EEG (electroencephalogram) signal generated by motor imagery
Figure BDA0002476162310000025
Obtained phase space vector
Figure BDA0002476162310000031
Number of each phase space vector
Figure BDA0002476162310000032
After the elements in (1) are sorted according to the amplitudeMapping to a symbol pic,uL- (d-1) tau symbols are obtained, including d! Different kinds of symbols, the set of all kinds of symbols being
Figure BDA0002476162310000033
Wherein r represents the r-th symbol, C ═ C3, C4, F3, F4, P3, P4, T7, T8, Cz; each phase space vector
Figure BDA0002476162310000034
Mapped symbol pic,uAll belonging to a collection of symbols of all kinds
Figure BDA0002476162310000035
Namely, it is
Figure BDA0002476162310000036
Weighted permutation entropy HWPECalculated by the following formula:
Figure BDA0002476162310000037
wherein, ω isc,uIs a phase space vector
Figure BDA0002476162310000038
Weight of pω(πc,r) Is the symbol pic,rWeighted likelihood of each phase space vector
Figure BDA0002476162310000039
Weight ω of (d)c,uCalculated by the following formula:
Figure BDA00024761623100000310
wherein, Xc,u+(m-1)τRepresenting phase space vectors
Figure BDA00024761623100000311
The m-th element of (a) is,
Figure BDA00024761623100000312
representing phase space vectors
Figure BDA00024761623100000313
The variance of (a);
(2) the wavelet packet energy entropy S2 is calculated as follows: motor imagery EEG electroencephalogram signal by wavelet packet decomposition
Figure BDA00024761623100000314
Decomposed into f levels with 2 at the f levelfA frequency band of L/2 for each frequency band of the f-th stagefWavelet packet coefficient, the expression of wavelet packet decomposition is:
Figure BDA00024761623100000315
wherein the content of the first and second substances,
Figure BDA00024761623100000316
indicating the kth wavelet packet coefficient at the η th frequency band at level f,
Figure BDA00024761623100000317
in order to be a function of the scale,
Figure BDA00024761623100000318
as a function of wavelets, the entropy of the energy of the wavelet packet HWPEECalculated by the following formula:
Figure BDA00024761623100000319
wherein p isηIs the probability of the energy of the η th frequency band, the wavelet packet decomposition uses the Daubechies 4 wavelet base (db4) to decompose into 5 layers, i.e. f is 5.
(3) The amplitude information plays an important role in revealing system dynamics, and the amplitude entropy S3 calculation process is as follows: firstly, the motor imagery EEG electroencephalogram signal is
Figure BDA00024761623100000320
The amplitude range of the EEG is divided into β intervals, and the EEG electroencephalogram signals are obtained by motor imagery
Figure BDA00024761623100000321
Is in each intervalγComprises the following steps:
Figure BDA00024761623100000322
wherein N isγIs the number of sample points whose amplitude falls within the gamma-th interval. Amplitude entropy HAECalculated by the following formula:
HAE＝-∑γpγlnpγ
as the dispersion of the signal amplitude increases, the amplitude entropy HAEAnd is increased.
The construction of the multi-entropy complex network in the step 4) comprises the following steps:
taking a channel corresponding to each electrode of the motor imagery EEG electroencephalogram as a node, calculating a two-norm distance between feature vectors corresponding to every two nodes, determining a threshold value by adopting a sparsity method, wherein the sparsity value is selected to be 20%, and if the two-norm distance between two nodes is smaller than the threshold value, a connecting edge exists between the two nodes to obtain a multi-entropy complex network; if a connecting edge exists between two nodes of the multi-entropy complex network, the position value of the two nodes corresponding to the adjacent matrix of the multi-entropy complex network is 1, and if the connecting edge does not exist between the two nodes of the multi-entropy complex network, the position value of the two nodes corresponding to the adjacent matrix of the multi-entropy complex network is 0. The dimensionality of an adjacent matrix of the multi-entropy complex network is Q multiplied by Q, Q is the number of nodes and is equal to the number of electrodes, namely Q is 9;
calculating a two-norm distance Rκ,νThen, the feature vector of node κ is set to [ S ]κ,1,Sκ,2,Sκ,3]Characteristic vector [ S ] of node vν,1,Sν,2,Sν,3]Calculated by the following formula:
Figure BDA0002476162310000041
where Q is the number of nodes, equal to the number of electrodes, i.e., Q9.
The network structure of the graph convolution neural network in the step 5) comprises four graph convolution layers, two graph pooling layers and a full connection layer, wherein the two graph convolution layers and the one graph pooling layer are sequentially connected to form a graph convolution module to form two graph convolution modules which are sequentially connected, and the output of the latter graph convolution module is the input of the full connection layer; each graph convolution layer is represented by the following nonlinear mapping function:
Hl+1＝σ(AHlWl)
wherein HlFor the first graph convolution layer feature, A is a multi-entropy complex network adjacency matrix, WlIs a parameter matrix of the ith graph convolution layer, sigma (-) is an activation function, and a ReLU function is adopted;
and extracting the characteristics capable of carrying out category distinguishing in the EEG signals of the motor imagery through four times of image convolution and two times of pooling operation, and finally flattening the characteristics and inputting the flattened characteristics into a full connection layer for carrying out motor imagery category identification.
Step 6) comprises the following steps:
(1) alternately playing hand fist making and stretching action videos, watching the videos by a testee and carrying out motor imagery of corresponding actions, and acquiring EEG electroencephalogram signals of the testee by an electroencephalogram signal acquisition device;
(2) the motor intention identification module preprocesses the motor imagery EEG electroencephalogram signals, constructs a multi-entropy complex network, inputs the multi-entropy complex network adjacent matrix into the graph convolution neural network to extract the signal characteristics and signal identification, obtains the motor intention of the testee and transmits the motor intention to the main controller;
(3) the myoelectric signal acquisition and multi-channel electrical stimulation output module is used for acquiring myoelectric signals of upper limb muscles of a tested person and transmitting the myoelectric signals to the main controller;
(4) the main controller obtains the movement intention of the testee according to the movement intention identification module, decides stimulation current, stimulation pulse width, stimulation frequency and stimulation time by combining the myoelectric signals of the upper limb muscles, and controls a plurality of electric stimulation points in the myoelectric signal acquisition and multi-channel electric stimulation output module;
(5) according to the instruction of the main controller, the plurality of electrical stimulation points apply electrical stimulation to a plurality of muscles of the upper limb corresponding to the movement intention, so that the testee can make hand fist making or stretching movement according to the imagination intention.
The brain control rehabilitation system motor imagery recognition system fusing the complex network and the graph convolution carries out feature extraction and recognition on the fist-making motor imagery EEG (electroencephalogram) and the hand stretching motor imagery EEG through constructing the multi-entropy complex network and further inputting the multi-entropy complex network into the graph convolution neural network, and converts the features into instructions to control the brain control rehabilitation system, helps patients with limited activity due to damage of upper limb motor pathways to carry out rehabilitation training, gradually recovers the activity of the patients, and promotes application of BCI (brain-computer interface) based on motor imagery in rehabilitation medicine.
Drawings
FIG. 1 is a flow chart of an exercise intention identification module of the present invention;
FIG. 2 is a timing diagram of single motor imagery EEG electroencephalogram acquisition in the present invention;
FIG. 3 is an overall block diagram of the brain-controlled rehabilitation system motor imagery identification system of the present invention, which integrates complex networks and graph convolutions;
FIG. 4 is a block diagram of a portable electroencephalogram acquisition device in the present invention;
fig. 5 is a schematic structural diagram of a functional electrical stimulation apparatus for limbs according to the present invention;
FIG. 6 is a schematic diagram of an electromyographic signal acquisition and multi-channel electrical stimulation output module according to the present invention.
In the drawings
1: the electroencephalogram signal acquisition device 2: motion intention recognition module
3: brain-controlled intelligent rehabilitation system 3.1: main controller
3.2: electromyographic signal acquisition and multi-channel electrical stimulation output module
3.2.1: electrical stimulation points 3.2.2: electromyographic signal acquisition point
Detailed Description
The brain control rehabilitation system motor imagery identification system fusing the complex network and the graph convolution is explained in detail below with reference to the embodiment and the accompanying drawings.
As shown in fig. 3, in the brain-controlled rehabilitation system motor imagery recognition system fusing a complex network and graph convolution, a human subject performs motor imagery by watching hand fist making and stretching action videos, and meanwhile, portable electroencephalogram signal acquisition equipment acquires motor imagery EEG electroencephalogram signals of the human subject; the motor intention identification module preprocesses the obtained motor imagery EEG electroencephalogram signals to construct a multi-entropy complex network, the multi-entropy complex network can fuse multi-channel motor imagery EEG electroencephalogram signals and extract the characteristics of the motor imagery EEG electroencephalogram signals in the aspects of symbol fluctuation, frequency energy distribution and amplitude fluctuation; the motor intention identification module inputs an adjacent matrix of the multi-entropy complex network into the atlas neural network to classify and identify the fist making motor imagery EEG electroencephalogram and the hand stretching motor imagery EEG electroencephalogram, and transmits the classification result to the brain control rehabilitation system to prompt the testee to execute the actions of making a fist and stretching the hand; the brain control rehabilitation system enables a closed-loop path to be formed between the movement intention and the body sensation of the limbs, gradually enhances the muscle strength and the nerve conduction speed of a testee, promotes the recovery of the movement area of the damaged brain, and finally realizes that the patient obtains the damaged movement function again through the autonomous training.
The portable electroencephalogram acquisition equipment adopts a structure disclosed by a patent application with the application number of 201810168228.X and the invention name of portable electroencephalogram acquisition equipment and application thereof in SSVEP and motor imagery, and comprises a system power supply circuit 11, a brain electrode cap transfer wire 12, a PGA amplification circuit 13, an AD converter 14, an STM32 processor 15 and a WIFI module 16, wherein the input end of the brain electrode cap transfer wire 12 is connected with a brain electrode cap for acquiring electroencephalogram signals, the output end of the brain electrode cap transfer wire is sequentially connected with the PGA amplification circuit 13, the AD converter 4 and the STM32 processor 15, the STM32 processor 15 is respectively connected with the PGA amplification circuit 13 and the AD converter 14 for controlling the working states of the PGA amplification circuit 13 and the AD converter 14, the WIFI module 16 is connected with the STM32 processor 15 for enabling the STM32 processor 15 to communicate with an upper computer through a wireless local area network, the system power supply circuit 11 is respectively connected with the PGA amplifying circuit 13, the AD converter 14, the STM32 processor 15 and the WIFI module 16 for supplying power. The brain electrode cap is an electrode cap.
As shown in fig. 3, the brain-controlled intelligent rehabilitation system of the present invention includes: the upper limb rehabilitation training device comprises a main controller 3.1 connected with the movement intention recognition module, and a myoelectric signal acquisition and multi-channel electrical stimulation output module 3.2 connected with the main controller 3.1, wherein the myoelectric signal acquisition and multi-channel electrical stimulation output module 3.2 applies electrical stimulation to the upper limb of a tested person according to a control command of the main controller 3.1 and feeds the myoelectric signal after electrical stimulation back to the main controller 3.1, and the main controller 3.1 generates a control command according to a classification result received from the movement intention recognition module and the myoelectric signal after electrical stimulation and controls the myoelectric signal acquisition and multi-channel electrical stimulation output module 3.2 to stimulate the upper limb of the tested person so as to help the tested person to carry out upper limb rehabilitation training.
As shown in fig. 1, the exercise intention identifying module specifically includes the following steps:
1) the method comprises the steps that a testee respectively watches a hand fist making video and a hand stretching video, meanwhile, motor imagery is carried out on corresponding actions of the videos, the portable electroencephalogram acquisition equipment is used for acquiring an EEG (electroencephalogram) signal for the fist making motor imagery of the testee and an EEG signal for the hand stretching motor imagery of the testee, the EEG signals are collectively called as the EEG signal for the motor imagery, and the EEG signal for the motor imagery is preprocessed;
the collecting of the EEG signals of the fist-making motor imagery EEG and the hand-stretching motor imagery EEG of the testee is to collect the EEG signals of the motor imagery EEG of nine electrodes of C3, C4, F3, F4, P3, P4, T7, T8 and Cz through brain electrode caps distributed according to 10-20 international standard leads and to carry out preprocessing. Before the experiment begins, each tested person carries out imagination training of making a fist and stretching the hand for many times so as to improve the signal quality. During the experiment, the testee sits in front of the computer screen. A timing chart of single motor imagery EEG electroencephalogram signal acquisition is shown in figure 2, and at the beginning of an experiment, a black cross appears in the center of a screen and is accompanied by a short voice prompt. After two seconds, a video of hand fist making or stretching appears on the screen at random, the video dwell time is 4s, the corresponding action is imagined according to the hand action in the video, and the video disappears until t is 6 s. After a short break, the process is repeated for the next motor imagery task.
The preprocessing is to perform artifact removal, 8-30Hz band-pass filtering and average removal on the motor imagery EEG electroencephalogram to obtain the motor imagery EEG electroencephalogram signal which can be used for realizing motor imagery state identification:
Figure BDA0002476162310000061
wherein Xc,iAnd the number of sampling points in channels corresponding to the electrodes with equal length is L.
2) Respectively calculating weighted permutation entropy S for the preprocessed motor imagery EEG signals of each electrode1Wavelet packet energy entropy S2Sum amplitude entropy S3Three entropies, and the three entropies are normalized based on a min-max standardization method; the three entropies can be used for describing and extracting time sequence nonlinear characteristics from different angles, the weighted arrangement entropy can be used for describing complexity from a time sequence fluctuation mode, namely a symbol representation angle, and the wavelet packet energy entropy and the amplitude entropy can be used for extracting characteristics from frequency energy distribution and amplitude fluctuation angles respectively; wherein said separately calculating a weighted permutation entropy S1Wavelet packet energy entropy S2Sum amplitude entropy S3The three entropies are specifically as follows:
(1) the weighted permutation entropy S1 is calculated as follows: motor imagery EEG electroencephalogram
Figure BDA0002476162310000062
The time delay embedding expression of (1) is:
Figure BDA0002476162310000071
wherein, Xc,iRepresents the ith sampling point in the motor imagery EEG electroencephalogram signal acquired by the c electrode, and L represents the c electrodeThe number of sampling points in the corresponding channel,
Figure BDA0002476162310000072
representing the u-th phase space vector generated by the channel corresponding to the c electrode, d is an embedding dimension, d is 4, tau is a delay time, tau is 1, and L- (d-1) tau represents an EEG (electroencephalogram) signal generated by motor imagery
Figure BDA0002476162310000073
Obtained phase space vector
Figure BDA0002476162310000074
Number of each phase space vector
Figure BDA0002476162310000075
The elements in the table are mapped into a symbol pi after being sorted according to the amplitudec,uL- (d-1) tau symbols are obtained, including d! Different kinds of symbols, the set of all kinds of symbols being
Figure BDA0002476162310000076
Wherein r represents the r-th symbol, C ═ C3, C4, F3, F4, P3, P4, T7, T8, Cz; each phase space vector
Figure BDA0002476162310000077
Mapped symbol pic,uAll belonging to a collection of symbols of all kinds
Figure BDA0002476162310000078
Namely, it is
Figure BDA0002476162310000079
By a four-dimensional vector
Figure BDA00024761623100000710
For example, a phase space vector is illustrated
Figure BDA00024761623100000711
According to the magnitude of the amplitudeMapped into a symbol pi after sortingc,uThe process of (2): the magnitude ordering of the four-dimensional vector is a2＞a1＞a3＞a4Whereby the four-dimensional vector can be transformed according to the subscript into the symbol 2134, up to 4! I.e. 24 symbols. If the phase space vector has elements with the same amplitude, the element with the earlier appearance time is judged to be a relatively larger value, and the element with the later appearance time is judged to be a relatively smaller value;
weighted permutation entropy HWPECalculated by the following formula:
Figure BDA00024761623100000712
wherein, ω isc,uIs a phase space vector
Figure BDA00024761623100000713
Weight of pω(πc,r) Is the symbol pic,rWeighted likelihood of each phase space vector
Figure BDA00024761623100000714
Weight ω of (d)c,uCalculated by the following formula:
Figure BDA00024761623100000715
wherein, Xc,u+(m-1)τRepresenting phase space vectors
Figure BDA00024761623100000716
The m-th element of (a) is,
Figure BDA00024761623100000717
representing phase space vectors
Figure BDA00024761623100000718
The variance of (a);
(2) the wavelet packet energy entropy S2 is calculated as follows: motor imagery EEG electroencephalogram signal by wavelet packet decomposition
Figure BDA00024761623100000719
Decomposed into f levels with 2 at the f levelfA frequency band of L/2 for each frequency band of the f-th stagefWavelet packet coefficient, the expression of wavelet packet decomposition is:
Figure BDA00024761623100000720
wherein the content of the first and second substances,
Figure BDA00024761623100000721
indicating the kth wavelet packet coefficient at the η th frequency band at level f,
Figure BDA00024761623100000722
in order to be a function of the scale,
Figure BDA00024761623100000723
as a function of wavelets, the entropy of the energy of the wavelet packet HWPEECalculated by the following formula:
Figure BDA0002476162310000081
wherein p isηIs the probability of the energy of the η th frequency band, the wavelet packet decomposition uses the Daubechies 4 wavelet base (db4) to decompose into 5 layers, i.e. f is 5;
(3) the amplitude information plays an important role in revealing system dynamics, and the amplitude entropy S3 calculation process is as follows: firstly, the motor imagery EEG electroencephalogram signal is
Figure BDA0002476162310000082
The amplitude range of the EEG is divided into β intervals, and the EEG electroencephalogram signals are obtained by motor imagery
Figure BDA0002476162310000083
Is in each intervalγComprises the following steps:
Figure BDA0002476162310000084
wherein N isγIs the number of sample points whose amplitude falls within the gamma-th interval. Amplitude entropy HAECalculated by the following formula:
HAE＝-∑γpγlnpγ
as the dispersion of the signal amplitude increases, the amplitude entropy HAEAnd is increased.
3) Constructing the three entropies of the weighted permutation entropy S1, the wavelet packet energy entropy S2 and the amplitude entropy S3 in the step 2) into a feature vector [ S1,S2,S3]；
4) Constructing a multi-entropy complex network based on the feature vectors in the step 3); the construction of the multi-entropy complex network comprises the following steps:
taking a channel corresponding to each electrode of the motor imagery EEG electroencephalogram as a node, calculating a two-norm distance between feature vectors corresponding to every two nodes, determining a threshold value by adopting a sparsity method, wherein the sparsity value is selected to be 20%, and if the two-norm distance between two nodes is smaller than the threshold value, a connecting edge exists between the two nodes to obtain a multi-entropy complex network; if a connecting edge exists between two nodes of the multi-entropy complex network, the position value of the two nodes corresponding to the adjacent matrix of the multi-entropy complex network is 1, and if the connecting edge does not exist between the two nodes of the multi-entropy complex network, the position value of the two nodes corresponding to the adjacent matrix of the multi-entropy complex network is 0. The dimension of the adjacent matrix of the multi-entropy complex network is Q multiplied by Q, wherein Q is the number of nodes and is equal to the number of electrodes, and Q is 9.
Calculating a two-norm distance Rκ,νThen, the feature vector of node κ is set to [ S ]κ,1,Sκ,2,Sκ,3]Characteristic vector [ S ] of node vν,1,Sν,2,Sν,3]Calculated by the following formula:
Figure BDA0002476162310000085
where Q is the number of nodes, equal to the number of electrodes, i.e., Q9.
5) Respectively constructing a multi-entropy complex network for a fist-making motor imagery EEG electroencephalogram signal and a hand stretching motor imagery EEG electroencephalogram signal in each motor imagery EEG electroencephalogram signal of each testee, and sending an adjacent matrix of the multi-entropy complex network and a classification label thereof into a graph convolution neural network for feature learning and classification;
the network structure of the graph convolution neural network comprises four graph convolution layers, two graph pooling layers and a full connection layer, wherein the two graph convolution layers and the one graph pooling layer are sequentially connected to form a graph convolution module to form two graph convolution modules which are sequentially connected, and the output of the latter graph convolution module is the input of the full connection layer; each graph convolution layer is represented by the following nonlinear mapping function:
Hl+1＝σ(AHlWl)
wherein HlFor the first graph convolution layer feature, A is a multi-entropy complex network adjacency matrix, WlIs a parameter matrix of the ith graph convolution layer, sigma (-) is an activation function, and a ReLU function is adopted;
and extracting the characteristics capable of carrying out category distinguishing in the EEG signals of the motor imagery through four times of image convolution and two times of pooling operation, and finally flattening the characteristics and inputting the flattened characteristics into a full connection layer for carrying out motor imagery category identification.
6) Transmitting the classification result to a brain control rehabilitation system to stimulate the upper limbs of the testee, as shown in fig. 3, the method comprises the following steps:
(1) alternately playing hand fist making and stretching action videos, watching the videos by a testee and carrying out motor imagery of corresponding actions, and acquiring EEG electroencephalogram signals of the testee by an electroencephalogram signal acquisition device;
(2) the motor intention identification module preprocesses the motor imagery EEG electroencephalogram signals, constructs a multi-entropy complex network, inputs the multi-entropy complex network adjacent matrix into the graph convolution neural network to extract the signal characteristics and signal identification, obtains the motor intention of the testee and transmits the motor intention to the main controller;
(3) the myoelectric signal acquisition and multi-channel electrical stimulation output module is used for acquiring myoelectric signals of upper limb muscles of a tested person and transmitting the myoelectric signals to the main controller;
(4) the main controller obtains the movement intention of the testee according to the movement intention identification module, and decides stimulation current, stimulation pulse width, stimulation frequency and stimulation time by combining myoelectric signals of upper limb muscles, specifically, the set stimulation current is 5-140 mA, the stimulation pulse width is 50-500 us, the stimulation frequency is 10-100 Hz, and the main controller controls a plurality of electric stimulation points in the myoelectric signal acquisition and multi-channel electric stimulation output module;
(5) according to the instruction of the main controller, the plurality of electrical stimulation points apply electrical stimulation to a plurality of muscles of the upper limb corresponding to the movement intention, so that the testee can make hand fist making or stretching movement according to the imagination intention. The brain control rehabilitation system based on motor imagery enables a closed loop path to be formed between motor intention and limb body sensation, muscle strength and nerve conduction speed of a testee are gradually enhanced, recovery of an injured brain movement area is promoted, and finally the purpose that the testee regains the movement function of limbs through autonomous training is achieved.
The step is realized on the limb functional electrical stimulation device shown in figure 5, the device comprises a portable electroencephalogram acquisition device 1, an electromyogram signal acquisition and multi-channel electrical stimulation output module 3.2 for performing electrical stimulation on lower limbs and electromyogram signal acquisition, and a sliding rod is arranged at the upper part of a backrest of a seat 4, so that the height of the portable electroencephalogram acquisition device 1 can be adjusted; the structural schematic diagram of the electromyographic signal acquisition and multi-channel electrical stimulation output module is shown in fig. 6, and the structural schematic diagram comprises a plurality of electrical stimulation points 3.2.1 and electromyographic signal acquisition points 3.2.2.
When the upper limb rehabilitation training is carried out, a tested person sits on the seat, the electroencephalogram signal acquisition equipment is adjusted to be suitable for the height of a patient through adjusting the sliding rod, the electrode cap is worn on the head of the tested person, the upper limb of the tested person is placed in the myoelectric signal acquisition and multi-channel electrical stimulation output module installed on the seat, and the tested person completes the autonomous rehabilitation training of the upper limb through carrying out motor imagery.
The above description of the present invention and the embodiments is not limited thereto, and the description of the embodiments is only one of the implementation manners of the present invention, and any structure or embodiment similar to the technical solution without inventive design is within the protection scope of the present invention without departing from the inventive spirit of the present invention.

Claims (9)

1. A brain control rehabilitation system motor imagery recognition system fusing a complex network and a atlas is characterized in that a testee carries out motor imagery by watching hand fist making and stretching action videos, and meanwhile, electroencephalogram signal acquisition equipment acquires motor imagery EEG (electroencephalogram) signals of the testee; the motor intention identification module preprocesses the obtained motor imagery EEG electroencephalogram signals to construct a multi-entropy complex network, the multi-entropy complex network can fuse multi-channel motor imagery EEG electroencephalogram signals and extract the characteristics of the motor imagery EEG electroencephalogram signals in the aspects of symbol fluctuation, frequency energy distribution and amplitude fluctuation; the motor intention identification module inputs an adjacent matrix of the multi-entropy complex network into the atlas neural network to classify and identify the fist making motor imagery EEG electroencephalogram and the hand stretching motor imagery EEG electroencephalogram, and transmits the classification result to the brain control rehabilitation system to prompt the testee to execute the actions of making a fist and stretching the hand; the brain control rehabilitation system enables a closed-loop path to be formed between the movement intention and the body sensation of the limbs, gradually enhances the muscle strength and the nerve conduction speed of a testee, promotes the recovery of the movement area of the damaged brain, and finally realizes that the patient obtains the damaged movement function again through the autonomous training.
2. The brain-controlled rehabilitation system motor imagery recognition system fusing complex networks and graph convolutions as claimed in claim 1, wherein said brain-controlled rehabilitation system comprises: the upper limb rehabilitation training device comprises a main controller connected with the movement intention recognition module, and a myoelectric signal acquisition and multi-channel electrical stimulation output module connected with the main controller, wherein the myoelectric signal acquisition and multi-channel electrical stimulation output module applies electrical stimulation to the upper limb of a tested person according to a control instruction of the main controller, feeds the electrically stimulated myoelectric signal back to the main controller, and the main controller generates a control instruction according to a classification result received from the movement intention recognition module and the electrically stimulated myoelectric signal, controls the myoelectric signal acquisition and multi-channel electrical stimulation output module to stimulate the upper limb of the tested person, and helps the tested person to perform upper limb rehabilitation training.
3. The brain-controlled rehabilitation system motor imagery recognition system fusing complex networks and graph convolutions as claimed in claim 1, wherein the motor intention recognition module specifically comprises the steps of:
1) the method comprises the steps that a testee respectively watches a hand fist making video and a hand stretching video, meanwhile, motor imagery is carried out on corresponding actions of the videos, electroencephalogram signal acquisition equipment is used for acquiring an EEG (electroencephalogram) electroencephalogram signal of the testee for making a fist making motor imagery and an EEG electroencephalogram signal of the hand stretching motor imagery, the EEG electroencephalogram signals are collectively called as the EEG electroencephalogram signals of the motor imagery, and the EEG electroencephalogram signals of the motor imagery are preprocessed;
2) respectively calculating weighted permutation entropy S for the preprocessed motor imagery EEG signals of each electrode1Wavelet packet energy entropy S2Sum amplitude entropy S3Three entropies, and the three entropies are normalized based on a min-max standardization method;
3) subjecting the weighted permutation entropy S of the step 2)1Wavelet packet energy entropy S2Sum amplitude entropy S3These three entropies are constructed into a feature vector [ S ]1,S2,S3]；
4) Constructing a multi-entropy complex network based on the feature vectors in the step 3);
5) respectively constructing a multi-entropy complex network for a fist-making motor imagery EEG electroencephalogram signal and a hand stretching motor imagery EEG electroencephalogram signal in each motor imagery EEG electroencephalogram signal of each testee, and sending an adjacent matrix of the multi-entropy complex network and a classification label thereof into a graph convolution neural network for feature learning and classification;
6) and transmitting the classification result to a brain control rehabilitation system to stimulate the upper limbs of the testee.
4. The brain-controlled rehabilitation system motor imagery recognition system fusing complex networks and graph convolutions as claimed in claim 3, wherein the step 1) of collecting the test subject's hand-clenched motor imagery EEG brain electrical signals and hand-stretched motor imagery EEG brain electrical signals is to collect the motor imagery EEG brain electrical signals of nine electrodes C3, C4, F3, F4, P3, P4, T7, T8 and Cz through brain electrode caps distributed according to 10-20 international standard leads, and to perform pre-processing, wherein the pre-processing is to perform de-artifact, band-pass filtering of 8-30Hz and de-averaging on the motor imagery EEG brain electrical signals to obtain the motor imagery EEG brain electrical signals capable of realizing motor imagery state recognition:
Figure FDA0002476162300000021
wherein Xc,iAnd the number of sampling points in channels corresponding to the electrodes with equal length is L.
5. The brain-controlled rehabilitation system motor imagery recognition system fusing complex networks and graph convolutions according to claim 3, wherein the separately calculated weighted permutation entropy S of step 2)1Wavelet packet energy entropy S2Sum amplitude entropy S3The three entropies are specifically as follows:
(1) weighted permutation entropy S1The calculation method of (2) is as follows: motor imagery EEG electroencephalogram
Figure FDA0002476162300000022
The time delay embedding expression of (1) is:
Figure FDA0002476162300000023
wherein, Xc,iRepresents the ith sampling point in the motor imagery EEG electroencephalogram signal collected by the c electrode, L represents the number of sampling points in the channel corresponding to the c electrode,
Figure FDA0002476162300000024
representing the u-th phase space vector generated by the channel corresponding to the c electrode, d is an embedding dimension, d is 4, tau is a delay time, tau is 1, and L- (d-1) tau represents an EEG (electroencephalogram) signal generated by motor imagery
Figure FDA0002476162300000025
Obtained phase space vector
Figure FDA0002476162300000026
Number of each phase space vector
Figure FDA0002476162300000027
The elements in the table are mapped into a symbol pi after being sorted according to the amplitudec,uL- (d-1) tau symbols are obtained, including d! Different kinds of symbols, the set of all kinds of symbols being
Figure FDA0002476162300000028
Wherein r represents the r-th symbol, C ═ C3, C4, F3, F4, P3, P4, T7, T8, Cz; each phase space vector
Figure FDA0002476162300000029
Mapped symbol pic,uAll belonging to a collection of symbols of all kinds
Figure FDA00024761623000000210
Namely, it is
Figure FDA00024761623000000211
Weighted permutation entropy HWPECalculated by the following formula:
Figure FDA00024761623000000212
wherein, ω isc,uIs a phase space vector
Figure FDA00024761623000000213
Weight of pω(πc,r) Is the symbol pic,rWeighted likelihood of each phase space vector
Figure FDA00024761623000000214
Weight ω of (d)c,uCalculated by the following formula:
Figure FDA00024761623000000215
wherein, Xc,u+(m-1)τRepresenting phase space vectors
Figure FDA00024761623000000216
The m-th element of (a) is,
Figure FDA00024761623000000217
representing phase space vectors
Figure FDA00024761623000000218
The variance of (a);
(2) the wavelet packet energy entropy S2 is calculated as follows: motor imagery EEG electroencephalogram signal by wavelet packet decomposition
Figure FDA0002476162300000031
Decomposed into f levels with 2 at the f levelfA frequency band of L/2 for each frequency band of the f-th stagefWavelet packet coefficient, the expression of wavelet packet decomposition is:
Figure FDA0002476162300000032
wherein the content of the first and second substances,
Figure FDA0002476162300000033
indicating the kth wavelet packet coefficient at the η th frequency band at level f,
Figure FDA0002476162300000034
in order to be a function of the scale,
Figure FDA0002476162300000035
as a function of wavelets, the entropy of the energy of the wavelet packet HWPEECalculated by the following formula:
Figure FDA0002476162300000036
wherein p isηIs the probability of the energy of the η th frequency band, the wavelet packet decomposition uses the Daubechies 4 wavelet base (db4) to decompose into 5 layers, i.e. f is 5.
(3) The amplitude information plays an important role in revealing system dynamics, and the amplitude entropy S3 calculation process is as follows: firstly, the motor imagery EEG electroencephalogram signal is
Figure FDA0002476162300000037
The amplitude range of the EEG is divided into β intervals, and the EEG electroencephalogram signals are obtained by motor imagery
Figure FDA0002476162300000038
Is in each intervalγComprises the following steps:
Figure FDA0002476162300000039
wherein N isγIs the number of sample points whose amplitude falls within the gamma-th interval. Amplitude entropy HAECalculated by the following formula:
HAE＝-∑γpγlnpγ
as the dispersion of the signal amplitude increases, the amplitude entropy HAEAnd is increased.
6. The brain-controlled rehabilitation system motor imagery recognition system fusing complex networks and graph convolution according to claim 3, wherein the constructing of the multi-entropy complex network in step 4) comprises:
taking a channel corresponding to each electrode of the motor imagery EEG electroencephalogram as a node, calculating a two-norm distance between feature vectors corresponding to every two nodes, determining a threshold value by adopting a sparsity method, wherein the sparsity value is selected to be 20%, and if the two-norm distance between two nodes is smaller than the threshold value, a connecting edge exists between the two nodes to obtain a multi-entropy complex network; if a connecting edge exists between two nodes of the multi-entropy complex network, the position value of the two nodes corresponding to the adjacent matrix of the multi-entropy complex network is 1, and if the connecting edge does not exist between the two nodes of the multi-entropy complex network, the position value of the two nodes corresponding to the adjacent matrix of the multi-entropy complex network is 0. The dimension of the adjacent matrix of the multi-entropy complex network is Q multiplied by Q, wherein Q is the number of nodes and is equal to the number of electrodes, and Q is 9.
7. The brain-controlled rehabilitation system motor imagery recognition system fusing complex networks and graph convolutions according to claim 6, wherein the calculating a two-norm distance Rκ,νThen, the feature vector of node κ is set to [ S ]κ,1,Sκ,2,Sκ,3]Characteristic vector [ S ] of node vν,1,Sν,2,Sν,3]Calculated by the following formula:
Figure FDA00024761623000000310
where Q is the number of nodes, equal to the number of electrodes, i.e., Q9.
8. The brain-controlled rehabilitation system motor imagery recognition system fusing complex networks and graph convolution according to claim 3, wherein the network structure of the graph convolution neural network in step 5) includes four graph convolution layers, two graph pooling layers and a full connection layer, the two graph convolution layers and the one graph pooling layer are sequentially connected to form a graph convolution module, so as to form two sequentially connected graph convolution modules, and the output of the latter graph convolution module is the input of the full connection layer; each graph convolution layer is represented by the following nonlinear mapping function:
Hl+1＝σ(AHlWl)
wherein HlFor the first graph convolution layer feature, A is a multi-entropy complex network adjacency matrix, WlIs the first graph convolution layerThe parameter matrix of (1), sigma (·) is an activation function, and a ReLU function is adopted;
and extracting the characteristics capable of carrying out category distinguishing in the EEG signals of the motor imagery through four times of image convolution and two times of pooling operation, and finally flattening the characteristics and inputting the flattened characteristics into a full connection layer for carrying out motor imagery category identification.
9. The brain-controlled rehabilitation system motor imagery recognition system fusing complex networks and graph convolutions of claim 3, wherein step 6) comprises:
(1) alternately playing hand fist making and stretching action videos, watching the videos by a testee and carrying out motor imagery of corresponding actions, and acquiring EEG electroencephalogram signals of the testee by an electroencephalogram signal acquisition device;
(2) the motor intention identification module preprocesses the motor imagery EEG electroencephalogram signals, constructs a multi-entropy complex network, inputs the multi-entropy complex network adjacent matrix into the graph convolution neural network to extract the signal characteristics and signal identification, obtains the motor intention of the testee and transmits the motor intention to the main controller;
(3) the myoelectric signal acquisition and multi-channel electrical stimulation output module is used for acquiring myoelectric signals of upper limb muscles of a tested person and transmitting the myoelectric signals to the main controller;
(4) the main controller obtains the movement intention of the testee according to the movement intention identification module, decides stimulation current, stimulation pulse width, stimulation frequency and stimulation time by combining the myoelectric signals of the upper limb muscles, and controls a plurality of electric stimulation points in the myoelectric signal acquisition and multi-channel electric stimulation output module;
(5) according to the instruction of the main controller, the plurality of electrical stimulation points apply electrical stimulation to a plurality of muscles of the upper limb corresponding to the movement intention, so that the testee can make hand fist making or stretching movement according to the imagination intention.
CN202010364649.7A 2020-04-30 2020-04-30 Brain control rehabilitation system motor imagery recognition system fusing complex network and graph convolution Active CN111584027B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010364649.7A CN111584027B (en) 2020-04-30 2020-04-30 Brain control rehabilitation system motor imagery recognition system fusing complex network and graph convolution

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010364649.7A CN111584027B (en) 2020-04-30 2020-04-30 Brain control rehabilitation system motor imagery recognition system fusing complex network and graph convolution

Publications (2)

Publication Number Publication Date
CN111584027A true CN111584027A (en) 2020-08-25
CN111584027B CN111584027B (en) 2022-03-11

Family

ID=72124616

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010364649.7A Active CN111584027B (en) 2020-04-30 2020-04-30 Brain control rehabilitation system motor imagery recognition system fusing complex network and graph convolution

Country Status (1)

Country Link
CN (1) CN111584027B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113499524A (en) * 2021-07-23 2021-10-15 华南理工大学 Auxiliary rehabilitation training system using motor imagery electroencephalogram detection
CN114176920A (en) * 2021-12-20 2022-03-15 曲阜师范大学 Intelligent wheelchair based on electroencephalogram control

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030225340A1 (en) * 2002-05-30 2003-12-04 Collura Thomas F. Repetitive visual stimulation to EEG neurofeedback protocols
US20050131311A1 (en) * 2003-12-12 2005-06-16 Washington University Brain computer interface
US20180268296A1 (en) * 2016-06-02 2018-09-20 Tencent Technology (Shenzhen) Company Limited Machine learning-based network model building method and apparatus
CN108597584A (en) * 2018-03-06 2018-09-28 上海大学 In conjunction with the three stages brain control upper limb healing method of Steady State Visual Evoked Potential and Mental imagery
CN109243569A (en) * 2018-06-29 2019-01-18 苏州大学 The recognition methods of spontaneous phantom object displacement orientation is tested based on brain information identification
CN110400619A (en) * 2019-08-30 2019-11-01 上海大学 A kind of healing hand function training method based on surface electromyogram signal
CN110534180A (en) * 2019-08-20 2019-12-03 西安电子科技大学 The man-machine coadaptation Mental imagery brain machine interface system of deep learning and training method

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030225340A1 (en) * 2002-05-30 2003-12-04 Collura Thomas F. Repetitive visual stimulation to EEG neurofeedback protocols
US20050131311A1 (en) * 2003-12-12 2005-06-16 Washington University Brain computer interface
US20180268296A1 (en) * 2016-06-02 2018-09-20 Tencent Technology (Shenzhen) Company Limited Machine learning-based network model building method and apparatus
CN108597584A (en) * 2018-03-06 2018-09-28 上海大学 In conjunction with the three stages brain control upper limb healing method of Steady State Visual Evoked Potential and Mental imagery
CN109243569A (en) * 2018-06-29 2019-01-18 苏州大学 The recognition methods of spontaneous phantom object displacement orientation is tested based on brain information identification
CN110534180A (en) * 2019-08-20 2019-12-03 西安电子科技大学 The man-machine coadaptation Mental imagery brain machine interface system of deep learning and training method
CN110400619A (en) * 2019-08-30 2019-11-01 上海大学 A kind of healing hand function training method based on surface electromyogram signal

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113499524A (en) * 2021-07-23 2021-10-15 华南理工大学 Auxiliary rehabilitation training system using motor imagery electroencephalogram detection
CN114176920A (en) * 2021-12-20 2022-03-15 曲阜师范大学 Intelligent wheelchair based on electroencephalogram control

Also Published As

Publication number Publication date
CN111584027B (en) 2022-03-11

Similar Documents

Publication Publication Date Title
CN110765920B (en) Motor imagery classification method based on convolutional neural network
CN111584030A (en) Idea control intelligent rehabilitation system based on deep learning and complex network and application
CN101711709B (en) Method for controlling electrically powered artificial hands by utilizing electro-coulogram and electroencephalogram information
CN111584031B (en) Brain-controlled intelligent limb rehabilitation system based on portable electroencephalogram acquisition equipment and application
CN110732082B (en) Exercise function rehabilitation method through transcranial direct current stimulation and functional electrical stimulation
CN110584660B (en) Electrode selection method based on brain source imaging and correlation analysis
CN111584027B (en) Brain control rehabilitation system motor imagery recognition system fusing complex network and graph convolution
CN110495893B (en) System and method for multi-level dynamic fusion recognition of continuous brain and muscle electricity of motor intention
CN104548347A (en) Pure idea nerve muscle electrical stimulation control and nerve function evaluation system
CN113398422A (en) Rehabilitation training system and method based on motor imagery-brain-computer interface and virtual reality
CN111584029B (en) Electroencephalogram self-adaptive model based on discriminant confrontation network and application of electroencephalogram self-adaptive model in rehabilitation
CN107822629B (en) Method for detecting myoelectric axes on surfaces of limbs
CN111616682A (en) Epileptic seizure early warning system based on portable electroencephalogram acquisition equipment and application
CN111616681B (en) Anesthesia state monitoring system based on portable electroencephalogram acquisition equipment and deep learning
CN104571504A (en) Online brain-machine interface method based on imaginary movement
CN107981997A (en) A kind of method for controlling intelligent wheelchair and system based on human brain motion intention
Awan et al. Effective classification of EEG signals using K-nearest neighbor algorithm
CN111631848A (en) Mind control artificial limb system based on brain-computer hybrid intelligence
CN111584032B (en) Novel brain-controlled intelligent rehabilitation method based on capsule network and application
Aljalal et al. Feature extraction of EEG based motor imagery using CSP based on logarithmic band power, entropy and energy
CN110472595B (en) Electroencephalogram recognition model construction method and device and recognition method and device
CN110262658B (en) Brain-computer interface character input system based on enhanced attention and implementation method
CN111584033B (en) Brain-controlled intelligent rehabilitation system movement intention recognition system based on multilayer ordered network
CN113647938A (en) Method and system for advanced detection of motion state change based on physiological signals
CN111584028B (en) Novel brain-controlled intelligent rehabilitation system based on visual graph symbol network and width learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant